WO2015169109A1 - A transformer noise suppression method - Google Patents

A transformer noise suppression method Download PDF

Info

Publication number
WO2015169109A1
WO2015169109A1 PCT/CN2015/070735 CN2015070735W WO2015169109A1 WO 2015169109 A1 WO2015169109 A1 WO 2015169109A1 CN 2015070735 W CN2015070735 W CN 2015070735W WO 2015169109 A1 WO2015169109 A1 WO 2015169109A1
Authority
WO
WIPO (PCT)
Prior art keywords
noise
particle
transformer
neural network
digital signal
Prior art date
Application number
PCT/CN2015/070735
Other languages
French (fr)
Chinese (zh)
Inventor
姜宁
马宏忠
李凯
许洪华
王春宁
陈冰冰
姜鸿羽
Original Assignee
国家电网公司
江苏省电力公司
江苏省电力公司南京供电公司
河海大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 国家电网公司, 江苏省电力公司, 江苏省电力公司南京供电公司, 河海大学 filed Critical 国家电网公司
Priority to US15/323,407 priority Critical patent/US10083709B2/en
Publication of WO2015169109A1 publication Critical patent/WO2015169109A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K11/00—Methods or devices for transmitting, conducting or directing sound in general; Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/16—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/175—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound
    • G10K11/178—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase
    • G10K11/1781—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase characterised by the analysis of input or output signals, e.g. frequency range, modes, transfer functions
    • G10K11/17821—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase characterised by the analysis of input or output signals, e.g. frequency range, modes, transfer functions characterised by the analysis of the input signals only
    • G10K11/17825—Error signals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K11/00—Methods or devices for transmitting, conducting or directing sound in general; Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/16—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/175—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound
    • G10K11/178—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase
    • G10K11/1785—Methods, e.g. algorithms; Devices
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K11/00—Methods or devices for transmitting, conducting or directing sound in general; Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/16—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/175—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound
    • G10K11/178—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase
    • G10K11/1785—Methods, e.g. algorithms; Devices
    • G10K11/17853—Methods, e.g. algorithms; Devices of the filter
    • G10K11/17854—Methods, e.g. algorithms; Devices of the filter the filter being an adaptive filter
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K11/00—Methods or devices for transmitting, conducting or directing sound in general; Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/16—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/175—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound
    • G10K11/178—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase
    • G10K11/1785—Methods, e.g. algorithms; Devices
    • G10K11/17855—Methods, e.g. algorithms; Devices for improving speed or power requirements
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K11/00—Methods or devices for transmitting, conducting or directing sound in general; Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/16—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
    • G10K11/175—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound
    • G10K11/178—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase
    • G10K11/1787—General system configurations
    • G10K11/17879—General system configurations using both a reference signal and an error signal
    • G10K11/17881—General system configurations using both a reference signal and an error signal the reference signal being an acoustic signal, e.g. recorded with a microphone
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G10L21/0232—Processing in the frequency domain
    • H—ELECTRICITY
    • H01—ELECTRIC ELEMENTS
    • H01F—MAGNETS; INDUCTANCES; TRANSFORMERS; SELECTION OF MATERIALS FOR THEIR MAGNETIC PROPERTIES
    • H01F27/00—Details of transformers or inductances, in general
    • H01F27/33—Arrangements for noise damping
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • H04R3/005—Circuits for transducers, loudspeakers or microphones for combining the signals of two or more microphones
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • H04R3/04—Circuits for transducers, loudspeakers or microphones for correcting frequency response
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K2210/00—Details of active noise control [ANC] covered by G10K11/178 but not provided for in any of its subgroups
    • G10K2210/10—Applications
    • G10K2210/125—Transformers
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K2210/00—Details of active noise control [ANC] covered by G10K11/178 but not provided for in any of its subgroups
    • G10K2210/30—Means
    • G10K2210/301—Computational
    • G10K2210/3026—Feedback
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K2210/00—Details of active noise control [ANC] covered by G10K11/178 but not provided for in any of its subgroups
    • G10K2210/30—Means
    • G10K2210/301—Computational
    • G10K2210/3027—Feedforward
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
    • G10K2210/00—Details of active noise control [ANC] covered by G10K11/178 but not provided for in any of its subgroups
    • G10K2210/30—Means
    • G10K2210/301—Computational
    • G10K2210/3038—Neural networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R2410/00—Microphones
    • H04R2410/05—Noise reduction with a separate noise microphone

Definitions

  • the invention relates to a transformer noise suppression method, in particular to a transformer noise suppression method based on an improved particle BP neural network, belonging to the technical field of power supply.
  • the noise of the transformer body is mainly caused by the electromagnetic vibration generated by the magnetostriction caused by the magnetostriction of the silicon steel sheet and the electromagnetic force generated by the current passing through the winding.
  • the energy of the transformer is mainly concentrated at the fundamental frequency (100 Hz) integral frequency below 500 Hz, and has strong penetrating power.
  • the characteristics of slow decay, etc.; the noise of the transformer cooling system mainly comes from the vibration of the fan and the oil pump during operation.
  • the energy is mainly concentrated in the high frequency part above 500Hz, and the noise attenuation is faster.
  • the low frequency noise of the transformer is the main cause of environmental pollution, and it is necessary to control this part of the noise.
  • Chinese patent CN 102355233A uses a direct digital synthesis of a fixed frequency sine and cosine signal as a reference signal, the frequency value of which corresponds to the nominal frequency of harmonic noise.
  • CN 102176668A directly synthesizes a corresponding reference signal based on picking up the fundamental frequency components in the primary noise. All of the above patents use digital direct synthesis of reference signals, which only eliminates noise from certain frequency components.
  • the traditional F-XLMS filtering algorithm can suppress all frequency component noise, but it can't deal with time-varying and nonlinear noise
  • scholars apply BP neural network algorithm with nonlinear processing ability, self-adaptive ability and robustness to the elimination.
  • the traditional BP neural network inherently has slow convergence speed and easy to fall into local minimum values, which makes the noise control effect not ideal. Therefore, it is necessary to find a filtering algorithm that can achieve better noise reduction effects.
  • the primary object of the present invention is to propose a transformer noise suppression method that can effectively eliminate noise.
  • a further object of the present invention is to provide a transformer noise suppression method based on an improved particle BP neural network that can overcome the above-mentioned prior art BP neural network with slow convergence speed and easy to fall into local minimum value defects.
  • the transformer single-channel noise suppression method of the present invention is mainly composed of a controller including a smart chip and an initial noise measuring microphone, a residual noise measuring microphone, and a speaker, wherein the controller implements noise according to the following steps. Suppression:
  • the first step is to receive an initial noise digital signal transmitted by the initial noise measurement microphone near the noise source and converted, as an input signal of a BP (Back Propagation) neural network;
  • BP Back Propagation
  • the secondary digital signal is converted into an analog signal, which is amplified and output to the speaker, and generates secondary noise that has an inhibitory effect on the initial noise;
  • the fourth step receiving the residual noise and the secondary noise superimposed by the residual noise measurement microphone and the converted residual noise digital signal, determining whether the amplitude of the residual noise digital signal has been continuously set the number of times, and if so, maintaining the secondary Digital signal output, if otherwise proceeding to the next step;
  • the BP neural network parameters are optimized and adjusted to minimize the amplitude of the residual noise digital signal, and the residual noise digital signal of the previous step is used as the input signal to repeat the second step.
  • the BP neural network parameters are optimized and adjusted by using an improved particle swarm algorithm, and the following process is performed:
  • Step 1 Determine the dimension of the particle according to the existing BP neural network structure, and randomly generate N initial particles;
  • Step 2 The weight coefficient ⁇ hi (n) between the input layer neuron i and the hidden layer neuron h in the neural network at the nth moment, and the weight coefficient W h between the hidden layer neuron h and the output layer ( n), the threshold GE h (n) of the hidden layer neuron h and the threshold ge(n) of the output layer neuron are real-coded in a predetermined order to form corresponding real numbers, each particle corresponding to a set of network parameters, Its coding form is;
  • Step 3 Select the residual noise signal e(n) of the system as the criterion for evaluating the network parameters, and use the following fitness function F(n) as the particle position coding and formula:
  • x(n) is the initial noise digital signal input at the time n;
  • H 1 (z) and H 2 (z) are the transfer functions of the primary channel and the secondary channel, respectively;
  • P g [p g1 ,p g2 ,...,p g(J(K+2)+1) ], otherwise, remain unchanged;
  • Step 5 If the number of iterations has reached the maximum number of iterations, the iteration is stopped, and the optimal position of the particle group is decoded to obtain the corresponding BP neural network parameters, otherwise the next step is performed;
  • Step 6 Define the evolution ⁇ of the particle swarm as:
  • f gbest (k) and f avg (k) are the global optimal fitness value and the average fitness value of the particle group at the kth iteration, respectively;
  • w 0 is the set initial value.
  • the formula for calculating the adaptive mutation probability is:
  • p m (k) is the mutation probability of the particle swarm at the kth iteration
  • p mmax and p mmin are the maximum and minimum of the mutation rate, respectively
  • f i (k) is the adaptation of the kth iteration of the particle i Degree value
  • ⁇ is the set constant
  • the new velocity and position calculation formula for defining particle variability is:
  • v id (k+1) w(k)v id (k)+c 1 r 1 (p id (k)-x id (k))+c 2 r 2 (p gd (k)-x id ( k)) (7)
  • v id (k) is the velocity of the d- th dimension of the particle i in the kth iteration
  • V i (k) is the velocity vector of the particle i in the kth iteration
  • p id (k) is the particle i k The best position of the dth dimension in the iteration
  • p gd (k) is the best position of the dth dimension in the kth iteration of the whole group
  • X i (k) is the position vector in the kth iteration of the particle i
  • c 1 And c 2 are non-negative acceleration constants
  • r 1 and r 2 are random numbers transformed in the range [0, 1]
  • R 1 and R 2 are unequal positive integers, in the range [1, N];
  • Step 7 First calculate the inertia factor and particle variability of this iteration from the above formulas (4) and (5), and then update the velocity and position of all particles by using equations (7) and (8) to generate a new generation of particles. Group, return to step four.
  • the digital signal x(n) at the nth time is input to the controller, it is first filtered by a BP neural network with random parameters to obtain an output value y(n), which is transmitted through the secondary acoustic channel. After the function is applied, the output value becomes s(n), that is, the secondary digital signal at the nth time, after superposition, the generated error digital signal is e(n).
  • the improved particle swarm optimization algorithm is used to correct the relevant parameters in the BP neural network in combination with the reference digital signal x(n) at the nth time.
  • the new weight coefficient and the threshold coefficient corresponding to the BP neural network at the nth time can be obtained, and then the new parameters are replaced with the original parameters.
  • the reference digital signal x(n+1) at the n+1th time is input to the controller, it is filtered by a BP neural network with updated parameters to obtain a new output value, which is subjected to a secondary acoustic channel transfer function. The output value then becomes s(n+1), which is the secondary digital signal at the n+1th time.
  • the invention not only establishes a complete single-channel transformer noise reduction system, but also can effectively realize noise reduction of noise sources such as transformers in the substation, and through scientific operation processing and multiple sets of single channel drops.
  • the coordination of the noise system can quickly suppress noise within a certain range of noise sources such as transformers, achieve an ideal noise reduction effect, and solve the problems of high cost, poor performance, and poor portability of the conventional noise reduction method.
  • FIG. 1 is a schematic diagram of a single channel noise reduction structure according to an embodiment of the present invention.
  • FIG. 2 is a flow chart showing the basic processing of the digital signal of the embodiment of FIG. 1.
  • FIG. 3 is a block diagram of an improved particle BP neural network system of the embodiment of FIG. 1.
  • FIG. 4 is a comparison diagram of the time domain effect of the residual noise measurement microphone pickup signal before and after noise reduction in the embodiment of FIG. 1.
  • FIG. 5 is a frequency domain effect diagram of the residual noise measurement microphone pickup signal before noise reduction in the embodiment of FIG. 1.
  • FIG. 6 is a frequency domain effect diagram of a residual noise measurement microphone pickup signal after noise reduction in the embodiment of FIG. 1.
  • Figure 7 is a multi-channel noise reduction system constructed in the embodiment of Figure 1.
  • This embodiment is a transformer noise suppression method based on the improved particle BP neural network.
  • the main components of the single channel noise reduction system include: a controller including a smart chip type TMS320VC5509 DSP chip, respectively as an initial noise measurement microphone And 2 PCM6110 microphones for residual noise measurement microphones, Swans S6.5 woofer.
  • the initial noise measuring microphone and the residual noise measuring microphone are respectively connected to the input port corresponding to the DSP chip of the controller through the sound processing chip (TLV320AIC23 audio chip) and the A/D conversion chip of the automatic gain adjuster, and the output end of the DSP chip passes through The D/A converter chip and power amplifier are connected to the speaker.
  • the signal acquisition frequency is set to 5000Hz
  • the BP neural network structure is set to 2-10-1 filtering algorithm.
  • the excitation functions of the hidden layer and the output layer are respectively Sigmoid function and Purelin function.
  • the primary noise measurement microphone is 10 cm from the transformer, the speaker is 30 cm from the transformer, and the residual noise measurement microphone is 20 cm from the speaker.
  • the controller When working, the controller implements noise suppression as follows (see Figure 2):
  • the first step is to receive an initial noise digital signal transmitted by the initial noise measurement microphone near the noise source and converted, as an input signal of a BP (Back Propagation) neural network;
  • BP Back Propagation
  • the secondary digital signal is converted into an analog signal, which is amplified and output to the speaker, and generates secondary noise that has an inhibitory effect on the initial noise;
  • the fourth step receiving the residual noise and the secondary noise superimposed by the residual noise measurement microphone and the converted residual noise digital signal, determining whether the amplitude of the residual noise digital signal has been continuously set the number of times, and if so, maintaining the secondary Digital signal output, if otherwise proceeding to the next step;
  • the BP neural network parameters are optimized and adjusted to minimize the amplitude of the residual noise digital signal, and the residual noise digital signal of the previous step is used as the input signal to repeat the second step; wherein the BP neural network parameters are optimized and adjusted.
  • Step 1 Determine the dimension of the particle according to the existing BP neural network structure, and randomly generate N initial particles;
  • Step 2 The weight coefficient ⁇ hi (n) between the input layer neuron i and the hidden layer neuron h in the neural network at the nth moment, and the weight coefficient W h between the hidden layer neuron h and the output layer ( n), the threshold GE h (n) of the hidden layer neuron h and the threshold ge(n) of the output layer neuron are real-coded in a predetermined order to form corresponding real numbers, each particle corresponding to a set of network parameters, Its coding form is;
  • Step 3 Select the residual noise signal e(n) of the system as the criterion for evaluating the network parameters, and use the following fitness function F(n) as the particle position coding and formula:
  • x(n) is the initial noise digital signal input at the time n;
  • H 1 (z) and H 2 (z) are the transfer functions of the primary channel and the secondary channel, respectively;
  • P g [p g1 ,p g2 ,...,p g(J(K+2)+1) ], otherwise, remain unchanged;
  • Step 5 If the number of iterations has reached the maximum number of iterations, the iteration is stopped, and the optimal position of the particle group is decoded to obtain the corresponding BP neural network parameters, otherwise the next step is performed;
  • Step 6 Define the evolution ⁇ of the particle swarm as:
  • f gbest (k) and f avg (k) are the global optimal fitness value and the average fitness value of the particle group at the kth iteration, respectively;
  • w 0 is the set initial value.
  • the formula for calculating the adaptive mutation probability is:
  • p m (k) is the mutation probability of the particle swarm at the kth iteration
  • p mmax and p mmin are the maximum and minimum of the mutation rate, respectively
  • f i (k) is the adaptation of the kth iteration of particle i Degree value
  • ⁇ is the set constant
  • the new velocity and position calculation formula for defining particle variability is:
  • v id (k+1) w(k)v id (k)+c 1 r 1 (p id (k)-x id (k))+c 2 r 2 (p gd (k)-x id ( k)) (7)
  • v id (k) is the velocity of the d- th dimension of the particle i in the kth iteration
  • V i (k) is the velocity vector of the particle i in the kth iteration
  • p id (k) is the particle i k The best position of the dth dimension in the iteration
  • p gd (k) is the best position of the dth dimension in the kth iteration of the whole group
  • X i (k) is the position vector in the kth iteration of the particle i
  • c 1 And c 2 are non-negative acceleration constants
  • r 1 and r 2 are random numbers transformed in the range [0, 1]
  • R 1 and R 2 are unequal positive integers, in the range [1, N];
  • Step 7 First calculate the inertia factor and particle variability of this iteration from the above formulas (4) and (5), and then update the velocity and position of all particles by using equations (7) and (8) to generate a new generation of particles. Group, return to step four.
  • this embodiment not only constructs a reasonable single-channel transformer noise suppression system, but also combines the particle swarm optimization algorithm based on BP neural network filtering, so as to improve the weight and threshold of BP neural network by using improved particle swarm optimization algorithm.
  • the correction method overcomes the inherent defects of the algorithm such as gradient descent optimization, and can effectively suppress the noise of all frequency components of the transformer.
  • a multi-channel noise reduction system consisting of multiple sets of single-channel noise reduction systems can effectively control the transformer noise within the allowable range.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Signal Processing (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Power Engineering (AREA)
  • Algebra (AREA)
  • Otolaryngology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Human Computer Interaction (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Quality & Reliability (AREA)
  • Databases & Information Systems (AREA)
  • Operations Research (AREA)
  • Soundproofing, Sound Blocking, And Sound Damping (AREA)

Abstract

A transformer noise suppression method, belonging to the technical field of electric power supply. The noise suppression method of individual active noise reduction system comprises the steps that: (1) initial noise digital signals are received and converted to serve as input signals of a BP neural network; (2) the input signals are processed to generate secondary digital signals; (3) the secondary digital signals are output to a loudspeaker and secondary noise is generated; (4) remained noise digital signals obtained by overlapping the initial noise and the secondary noise are received; whether remained noise digital signals is continuously constant for the set times is judged; if yes, the secondary digital signals are kept outputting; (5)if not, BP neural network parameters are optimized and adjusted with the amplitude of the remained noise digital signals being minimum as the optimality principle; remained noise digital signals of previous step are served as new input signals and the step (2) is executed again. At proper position about oil boxes of transformers multi individual path noise reduction systems are arranged in order to constitute the multi-path suppression system of transformer noise. The transformer noise suppression method can quickly suppress the noise within a certain range of noise sources such as transformers; an ideal noise reduction effect can be achieved and the problems of high cost, poor effect and poor portability with traditional noise reduction methods are solved.

Description

一种变压器噪声抑制方法Transformer noise suppression method 技术领域Technical field
本发明涉及一种变压器噪声抑制方法，尤其是一种基于改进粒子BP神经网络的变压器噪声抑制方法，属于电力供应技术领域。The invention relates to a transformer noise suppression method, in particular to a transformer noise suppression method based on an improved particle BP neural network, belonging to the technical field of power supply.
背景技术Background technique
研究表明，变电站噪声主要为大型电力变压器噪声，而电力变压器噪声来源于变压器本体和冷却系统两个方面。Research shows that substation noise is mainly for large power transformer noise, and power transformer noise comes from two aspects of transformer body and cooling system.
变压器本体噪声主要源于硅钢片磁致伸缩所引起的铁心振动和电流通过绕组所产生的电磁力，其能量主要集中在500Hz以下的基频(100Hz)整数倍频率处，具有穿透能力强、衰减慢等特点；变压器冷却系统噪声主要源于风扇和油泵运行时的振动，其能量主要集中在500Hz以上的高频部分，该部分噪声衰减较快。变压器低频噪声是造成环境污染的主要因素，控制该部分噪声很有必要。The noise of the transformer body is mainly caused by the electromagnetic vibration generated by the magnetostriction caused by the magnetostriction of the silicon steel sheet and the electromagnetic force generated by the current passing through the winding. The energy of the transformer is mainly concentrated at the fundamental frequency (100 Hz) integral frequency below 500 Hz, and has strong penetrating power. The characteristics of slow decay, etc.; the noise of the transformer cooling system mainly comes from the vibration of the fan and the oil pump during operation. The energy is mainly concentrated in the high frequency part above 500Hz, and the noise attenuation is faster. The low frequency noise of the transformer is the main cause of environmental pollution, and it is necessary to control this part of the noise.
中国专利CN 102355233A采用直接数字合成固定频率的正余弦信号作为参考信号，其频率值对应谐波噪声的额定频率。CN 102176668A根据拾取初级噪声中的基波频率成分，直接数字合成相应的参考信号。以上专利均利用数字直接合成参考信号，只能消除某些频率成分的噪声。虽然传统的F-XLMS滤波算法可以抑制所有频率成分噪声，但是其无法处理时变、非线性噪声，故学者将具有非线性处理能力、自适应能力及鲁棒性的BP神经网络算法应用到消噪领域中，而传统BP神经网络固有的收敛速度慢和容易陷入局部极小值等缺陷，使其噪声控制的效果并不理想。因此，需寻求一种能够取得更好降噪效果的滤波算法。Chinese patent CN 102355233A uses a direct digital synthesis of a fixed frequency sine and cosine signal as a reference signal, the frequency value of which corresponds to the nominal frequency of harmonic noise. CN 102176668A directly synthesizes a corresponding reference signal based on picking up the fundamental frequency components in the primary noise. All of the above patents use digital direct synthesis of reference signals, which only eliminates noise from certain frequency components. Although the traditional F-XLMS filtering algorithm can suppress all frequency component noise, but it can't deal with time-varying and nonlinear noise, scholars apply BP neural network algorithm with nonlinear processing ability, self-adaptive ability and robustness to the elimination. In the noise field, the traditional BP neural network inherently has slow convergence speed and easy to fall into local minimum values, which makes the noise control effect not ideal. Therefore, it is necessary to find a filtering algorithm that can achieve better noise reduction effects.
发明内容Summary of the invention
本发明的首要目的在于：提出一种可以有效消除噪声的变压器噪声抑制方法。The primary object of the present invention is to propose a transformer noise suppression method that can effectively eliminate noise.
本发明进一步的目的：提出一种可以克服上述现有技术BP神经网络收敛速度慢和容易陷入局部极小值缺陷的基于改进粒子BP神经网络的变压器噪声抑制方法。A further object of the present invention is to provide a transformer noise suppression method based on an improved particle BP neural network that can overcome the above-mentioned prior art BP neural network with slow convergence speed and easy to fall into local minimum value defects.
为了达到以上首要目的，本发明的变压器单通道噪声抑制方法在主要由含智能芯片的控制器以及初始噪声测量传声器、残余噪声测量传声器、扬声器组成的系统中，所述控制器按以下步骤实现噪声抑制： In order to achieve the above primary object, the transformer single-channel noise suppression method of the present invention is mainly composed of a controller including a smart chip and an initial noise measuring microphone, a residual noise measuring microphone, and a speaker, wherein the controller implements noise according to the following steps. Suppression:
第一步、接收位于噪声源附近初始噪声测量传声器传输并经转换的初始噪声数字信号，作为BP(Back Propagation)神经网络的输入信号；The first step is to receive an initial noise digital signal transmitted by the initial noise measurement microphone near the noise source and converted, as an input signal of a BP (Back Propagation) neural network;
第二步、借助BP神经网络对输入信号进行处理后，生成相位与输入信号偏离的次级数字信号；In the second step, after the input signal is processed by the BP neural network, a secondary digital signal whose phase deviates from the input signal is generated;
第三步、将次级数字信号转换成模拟信号经放大后输出到扬声器，产生对初始噪声具有抑制作用的次级噪声；In the third step, the secondary digital signal is converted into an analog signal, which is amplified and output to the speaker, and generates secondary noise that has an inhibitory effect on the initial noise;
第四步、接收残余噪声测量传声器拾取的初始噪声和次级噪声叠加并经转换后的残余噪声数字信号，判断残余噪声数字信号的幅值是否已连续设定次数不变，如是则保持次级数字信号输出，如否则进行下步；The fourth step, receiving the residual noise and the secondary noise superimposed by the residual noise measurement microphone and the converted residual noise digital signal, determining whether the amplitude of the residual noise digital signal has been continuously set the number of times, and if so, maintaining the secondary Digital signal output, if otherwise proceeding to the next step;
第五步、以使残余噪声数字信号幅值最小为优化原则，优化、调整BP神经网络参数，将上一步的残余噪声数字信号作为输入信号重复第二步。In the fifth step, the BP neural network parameters are optimized and adjusted to minimize the amplitude of the residual noise digital signal, and the residual noise digital signal of the previous step is used as the input signal to repeat the second step.
为了达到进一步的目的，本发明变压器单通道噪声抑制方法中，所述第五步中优化、调整BP神经网络参数采用改进粒子群算法，按以下过程进行：In order to achieve a further object, in the single-channel noise suppression method of the transformer of the present invention, in the fifth step, the BP neural network parameters are optimized and adjusted by using an improved particle swarm algorithm, and the following process is performed:
步骤一、根据已有的BP神经网络结构确定粒子的维数，并随机产生N个初始粒子；Step 1: Determine the dimension of the particle according to the existing BP neural network structure, and randomly generate N initial particles;
步骤二、将第n时刻神经网络中输入层神经元i与隐含层神经元h之间的权系数ωhi(n)、隐含层神经元h与输出层之间的权系数Wh(n)、隐含层神经元h的阈值GEh(n)及输出层神经元的阈值ge(n)按预定的次序进行实数编码，形成相应的实数粒子，每个粒子对应一组网络参数，其编码形式为；Step 2: The weight coefficient ω hi (n) between the input layer neuron i and the hidden layer neuron h in the neural network at the nth moment, and the weight coefficient W h between the hidden layer neuron h and the output layer ( n), the threshold GE h (n) of the hidden layer neuron h and the threshold ge(n) of the output layer neuron are real-coded in a predetermined order to form corresponding real numbers, each particle corresponding to a set of network parameters, Its coding form is;
ω11 Ω 11 ω12 ω 12 …... ω1K ω 1K W1 W 1 GE1 GE 1 …... ωJ1 ω J1 ωJ2 ω J2 …... ωJK ω JK WJ W J GEJ GE J geGe
步骤三、选择系统的残余噪声信号e(n)作为网络参数的评判标准，以如下适应度函数F(n)作为粒子位置编码和公式：Step 3: Select the residual noise signal e(n) of the system as the criterion for evaluating the network parameters, and use the following fitness function F(n) as the particle position coding and formula:
Figure PCTCN2015070735-appb-000001
Figure PCTCN2015070735-appb-000001
上式中，x(n)为第n时刻输入的初始噪声数字信号；H1(z)和H2(z)分别为初级通道和次级通道的传递函数；In the above formula, x(n) is the initial noise digital signal input at the time n; H 1 (z) and H 2 (z) are the transfer functions of the primary channel and the secondary channel, respectively;
Figure PCTCN2015070735-appb-000002
Figure PCTCN2015070735-appb-000002
上式中，xi(n)＝x(n-i+1)表示输入层神经元i的输入；以f(x)＝2/(1+exp(-2x))-1表示网络隐含层的激励函数；In the above formula, x i (n)=x(n-i+1) represents the input of the input layer neuron i; the network implied by f(x)=2/(1+exp(-2x))-1 The excitation function of the layer;
步骤四、根据粒子位置编码和公式(1)，计算出各粒子的适应度值；将粒子当前位置适应度值与之前迭代时该粒子的位置适应度值作比较，若前者比后者小，则更新该粒子最优位置Pi＝[pi1,pi2,…,pi(J(K+2)+1)]，否则，保持不变；同时，将该粒子对应的适应度值与之前迭代时的适应度值作比较，若前者比后者小，则更新整个粒子群的最优位置Pg＝[pg1,pg2,…,pg(J(K+2)+1)]，否则，保持不变；Step 4: Calculate the fitness value of each particle according to the particle position coding and formula (1); compare the current position fitness value of the particle with the position fitness value of the particle in the previous iteration, if the former is smaller than the latter, Then updating the optimal position of the particle P i =[p i1 ,p i2 ,...,p i(J(K+2)+1) ], otherwise, remains unchanged; at the same time, the corresponding fitness value of the particle is The fitness values of the previous iterations are compared. If the former is smaller than the latter, the optimal position of the entire particle swarm is updated. P g =[p g1 ,p g2 ,...,p g(J(K+2)+1) ], otherwise, remain unchanged;
步骤五、若迭代次数已达到最大迭代次数，则停止迭代，对粒子群最优位置进行解码得到相应BP神经网络参数，否则进行下一步；Step 5: If the number of iterations has reached the maximum number of iterations, the iteration is stopped, and the optimal position of the particle group is decoded to obtain the corresponding BP neural network parameters, otherwise the next step is performed;
步骤六、定义粒子群的进化度σ为： Step 6. Define the evolution σ of the particle swarm as:
Figure PCTCN2015070735-appb-000003
Figure PCTCN2015070735-appb-000003
上式中，fgbest(k)和favg(k)分别为第k次迭代时粒子群的全局最优适应度值和平均适应度值；In the above formula, f gbest (k) and f avg (k) are the global optimal fitness value and the average fitness value of the particle group at the kth iteration, respectively;
定义动态变化的惯性因子计算公式为：The formula for calculating the inertia factor for dynamically changing is:
Figure PCTCN2015070735-appb-000004
Figure PCTCN2015070735-appb-000004
上式中，w0为设定的初始值。In the above formula, w 0 is the set initial value.
定义自适应变异概率计算公式为：The formula for calculating the adaptive mutation probability is:
Figure PCTCN2015070735-appb-000005
Figure PCTCN2015070735-appb-000005
Figure PCTCN2015070735-appb-000006
Figure PCTCN2015070735-appb-000006
其中，pm(k)为粒子群在第k次迭代的变异概率；pmmax和pmmin分别为变异率的 最大值和最小值，fi(k)为粒子i第k次迭代时的适应度值，ε为设定的常数；Where p m (k) is the mutation probability of the particle swarm at the kth iteration; p mmax and p mmin are the maximum and minimum of the mutation rate, respectively, and f i (k) is the adaptation of the kth iteration of the particle i Degree value, ε is the set constant;
定义粒子变异后新的速度和位置计算公式为：The new velocity and position calculation formula for defining particle variability is:
vid(k+1)＝w(k)vid(k)+c1r1(pid(k)-xid(k))+c2r2(pgd(k)-xid(k))   (7)v id (k+1)=w(k)v id (k)+c 1 r 1 (p id (k)-x id (k))+c 2 r 2 (p gd (k)-x id ( k)) (7)
Figure PCTCN2015070735-appb-000007
Figure PCTCN2015070735-appb-000007
其中，vid(k)为粒子i在第k次迭代中第d维的速度；Vi(k)为粒子i在第k次迭代中的速度向量；pid(k)为粒子i第k次迭代中第d维的最好位置；pgd(k)为整个群体第k次迭代中第d维的最好位置；Xi(k)为粒子i第k次迭代中位置向量；c1和c2为非负加速常数；r1和r2为[0,1]范围内变换的随机数；R1和R2为不相等正整数，在[1,N]范围内；Where v id (k) is the velocity of the d- th dimension of the particle i in the kth iteration; V i (k) is the velocity vector of the particle i in the kth iteration; p id (k) is the particle i k The best position of the dth dimension in the iteration; p gd (k) is the best position of the dth dimension in the kth iteration of the whole group; X i (k) is the position vector in the kth iteration of the particle i; c 1 And c 2 are non-negative acceleration constants; r 1 and r 2 are random numbers transformed in the range [0, 1]; R 1 and R 2 are unequal positive integers, in the range [1, N];
步骤七、先由以上公式(4)和(5)计算出本次迭代时的惯性因子和粒子变异率，再利用公式(7)和(8)更新所有粒子的速度和位置，生成新一代粒子群，返回步骤四。Step 7. First calculate the inertia factor and particle variability of this iteration from the above formulas (4) and (5), and then update the velocity and position of all particles by using equations (7) and (8) to generate a new generation of particles. Group, return to step four.
这样即可实现，当第n时刻的数字信号x(n)输入控制器时，先将其通过一个具有随机参数的BP神经网络滤波，得到一个输出值y(n)，经过次级声通道传递函数作用后该输出值变为s(n)，即第n时刻的次级数字信号，经过叠加后，产生的误差数字信号为e(n)。当第n时刻的误差数字信号e(n)反馈输入控制器时，结合第n时刻的参考数字信号x(n)，利用改进粒子群优化算法修正BP神经网络中相关参数。经过上述运算处理后，便可以得到第n时刻BP神经网络对应的新的权系数和阈系数，然后将这些新的参数替换原来的参数。当第n+1时刻的参考数字信号x(n+1)输入控制器时，便将其通过一个更新过参数的BP神经网络滤波，得到一个新的输出值，经过次级声通道传递函数作用后该输出值变为s(n+1)，即第n+1时刻的次级数字信号。In this way, when the digital signal x(n) at the nth time is input to the controller, it is first filtered by a BP neural network with random parameters to obtain an output value y(n), which is transmitted through the secondary acoustic channel. After the function is applied, the output value becomes s(n), that is, the secondary digital signal at the nth time, after superposition, the generated error digital signal is e(n). When the error digital signal e(n) at the nth time is fed back to the controller, the improved particle swarm optimization algorithm is used to correct the relevant parameters in the BP neural network in combination with the reference digital signal x(n) at the nth time. After the above operation processing, the new weight coefficient and the threshold coefficient corresponding to the BP neural network at the nth time can be obtained, and then the new parameters are replaced with the original parameters. When the reference digital signal x(n+1) at the n+1th time is input to the controller, it is filtered by a BP neural network with updated parameters to obtain a new output value, which is subjected to a secondary acoustic channel transfer function. The output value then becomes s(n+1), which is the secondary digital signal at the n+1th time.
本发明与现有技术相比，不仅由于建立了一套完整的单通道变压器降噪系统，可以切实实现对变电站内变压器等噪声源的降噪，而且通过科学的运算处理和多组单通道降噪系统的配合，能够迅速抑制变压器等噪声源一定范围内的噪声，取得理想的降噪效果，解决了传统降噪方法存在的费用高、效果差及可移植性差等问题。Compared with the prior art, the invention not only establishes a complete single-channel transformer noise reduction system, but also can effectively realize noise reduction of noise sources such as transformers in the substation, and through scientific operation processing and multiple sets of single channel drops. The coordination of the noise system can quickly suppress noise within a certain range of noise sources such as transformers, achieve an ideal noise reduction effect, and solve the problems of high cost, poor performance, and poor portability of the conventional noise reduction method.
附图说明 DRAWINGS
图1是本发明一个实施例的单通道降噪结构示意图。1 is a schematic diagram of a single channel noise reduction structure according to an embodiment of the present invention.
图2是图1实施例的数字信号基本处理过程流程图。2 is a flow chart showing the basic processing of the digital signal of the embodiment of FIG. 1.
图3是图1实施例的改进粒子BP神经网络系统框图。3 is a block diagram of an improved particle BP neural network system of the embodiment of FIG. 1.
图4是图1实施例降噪前后残余噪声测量传声器拾取信号的时域效果对照图。4 is a comparison diagram of the time domain effect of the residual noise measurement microphone pickup signal before and after noise reduction in the embodiment of FIG. 1.
图5是图1实施例降噪前残余噪声测量传声器拾取信号的频域效果图。FIG. 5 is a frequency domain effect diagram of the residual noise measurement microphone pickup signal before noise reduction in the embodiment of FIG. 1. FIG.
图6是图1实施例降噪后残余噪声测量传声器拾取信号的频域效果图。6 is a frequency domain effect diagram of a residual noise measurement microphone pickup signal after noise reduction in the embodiment of FIG. 1.
图7是图1实施例组成的多通道降噪系统。Figure 7 is a multi-channel noise reduction system constructed in the embodiment of Figure 1.
具体实施方式detailed description
实施例一 Embodiment 1
下面结合附图对本发明的技术方案做进一步的详细说明。The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.
本实施例为基于改进粒子BP神经网络的变压器噪声抑制方法，如图1所示，单通道降噪系统主要构成包括：含智能芯片——型号为TMS320VC5509DSP芯片的控制器，分别作为初始噪声测量传声器和残余噪声测量传声器的2只PCM6110麦克风，惠威S6.5低频扬声器。其中初始噪声测量传声器和残余噪声测量传声器分别经过自动增益调节器的声音处理芯片(TLV320AIC23音频芯片)和A/D转换芯片后与控制器的DSP芯片对应的输入端口连接，DSP芯片的输出端通过D/A转换芯片和功率放大器后接扬声器。具体针对某变压器现场，信号采集频率设定为5000Hz，BP神经网络结构设为2-10-1滤波算法，其中隐含层和输出层的激励函数分别为Sigmoid函数和Purelin函数。P粒子群的规模为30，最大遗传代数为10。根据粒子的编码方式，粒子的维数为78，c1＝c2＝1.3，ω0＝0.9，pmmax＝0.1，pmmin＝0.05。初级噪声测量传声器距离变压器10cm,扬声器距离变压器30cm，残余噪声测量传声器距离扬声器20cm。This embodiment is a transformer noise suppression method based on the improved particle BP neural network. As shown in FIG. 1 , the main components of the single channel noise reduction system include: a controller including a smart chip type TMS320VC5509 DSP chip, respectively as an initial noise measurement microphone And 2 PCM6110 microphones for residual noise measurement microphones, Swans S6.5 woofer. The initial noise measuring microphone and the residual noise measuring microphone are respectively connected to the input port corresponding to the DSP chip of the controller through the sound processing chip (TLV320AIC23 audio chip) and the A/D conversion chip of the automatic gain adjuster, and the output end of the DSP chip passes through The D/A converter chip and power amplifier are connected to the speaker. Specifically, for a transformer site, the signal acquisition frequency is set to 5000Hz, and the BP neural network structure is set to 2-10-1 filtering algorithm. The excitation functions of the hidden layer and the output layer are respectively Sigmoid function and Purelin function. The P particle population has a scale of 30 and a maximum genetic algebra of 10. According to the way the particles are encoded, the dimension of the particles is 78, c 1 = c 2 = 1.3, ω 0 = 0.9, p mmax = 0.1, p mmin = 0.05. The primary noise measurement microphone is 10 cm from the transformer, the speaker is 30 cm from the transformer, and the residual noise measurement microphone is 20 cm from the speaker.
工作时，控制器按以下步骤实现噪声抑制(参见图2)：When working, the controller implements noise suppression as follows (see Figure 2):
第一步、接收位于噪声源附近初始噪声测量传声器传输并经转换的初始噪声数字信号，作为BP(Back Propagation)神经网络的输入信号；The first step is to receive an initial noise digital signal transmitted by the initial noise measurement microphone near the noise source and converted, as an input signal of a BP (Back Propagation) neural network;
第二步、借助BP神经网络对输入信号进行滤波处理后，生成相位与输入信号偏离的次级数字信号；In the second step, after the input signal is filtered by the BP neural network, a secondary digital signal whose phase deviates from the input signal is generated;
第三步、将次级数字信号转换成模拟信号经放大后输出到扬声器，产生对初始噪声具有抑制作用的次级噪声； In the third step, the secondary digital signal is converted into an analog signal, which is amplified and output to the speaker, and generates secondary noise that has an inhibitory effect on the initial noise;
第四步、接收残余噪声测量传声器拾取的初始噪声和次级噪声叠加并经转换后的残余噪声数字信号，判断残余噪声数字信号的幅值是否已连续设定次数不变，如是则保持次级数字信号输出，如否则进行下步；The fourth step, receiving the residual noise and the secondary noise superimposed by the residual noise measurement microphone and the converted residual noise digital signal, determining whether the amplitude of the residual noise digital signal has been continuously set the number of times, and if so, maintaining the secondary Digital signal output, if otherwise proceeding to the next step;
第五步、以使残余噪声数字信号幅值最小为优化原则，优化、调整BP神经网络参数，将上一步的残余噪声数字信号作为输入信号重复第二步；其中优化、调整BP神经网络参数采用改进粒子群算法，按以下基本过程进行(参见图3)：In the fifth step, the BP neural network parameters are optimized and adjusted to minimize the amplitude of the residual noise digital signal, and the residual noise digital signal of the previous step is used as the input signal to repeat the second step; wherein the BP neural network parameters are optimized and adjusted. To improve the particle swarm algorithm, follow the basic process below (see Figure 3):
步骤一、根据已有的BP神经网络结构确定粒子的维数，并随机产生N个初始粒子；Step 1: Determine the dimension of the particle according to the existing BP neural network structure, and randomly generate N initial particles;
步骤二、将第n时刻神经网络中输入层神经元i与隐含层神经元h之间的权系数ωhi(n)、隐含层神经元h与输出层之间的权系数Wh(n)、隐含层神经元h的阈值GEh(n)及输出层神经元的阈值ge(n)按预定的次序进行实数编码，形成相应的实数粒子，每个粒子对应一组网络参数，其编码形式为；Step 2: The weight coefficient ω hi (n) between the input layer neuron i and the hidden layer neuron h in the neural network at the nth moment, and the weight coefficient W h between the hidden layer neuron h and the output layer ( n), the threshold GE h (n) of the hidden layer neuron h and the threshold ge(n) of the output layer neuron are real-coded in a predetermined order to form corresponding real numbers, each particle corresponding to a set of network parameters, Its coding form is;
ω11 Ω 11 ω12 ω 12 …... ω1K ω 1K W1 W 1 GE1 GE 1 …... ωJ1 ω J1 ωJ2 ω J2 …... ωJK ω JK WJ W J GEJ GE J geGe
步骤三、选择系统的残余噪声信号e(n)作为网络参数的评判标准，以如下适应度函数F(n)作为粒子位置编码和公式：Step 3: Select the residual noise signal e(n) of the system as the criterion for evaluating the network parameters, and use the following fitness function F(n) as the particle position coding and formula:
Figure PCTCN2015070735-appb-000008
Figure PCTCN2015070735-appb-000008
上式中，x(n)为第n时刻输入的初始噪声数字信号；H1(z)和H2(z)分别为初级通道和次级通道的传递函数；In the above formula, x(n) is the initial noise digital signal input at the time n; H 1 (z) and H 2 (z) are the transfer functions of the primary channel and the secondary channel, respectively;
Figure PCTCN2015070735-appb-000009
Figure PCTCN2015070735-appb-000009
上式中，xi(n)＝x(n-i+1)表示输入层神经元i的输入；以f(x)＝2/(1+exp(-2x))-1表示网络隐含层的激励函数；In the above formula, x i (n)=x(n-i+1) represents the input of the input layer neuron i; the network implied by f(x)=2/(1+exp(-2x))-1 The excitation function of the layer;
步骤四、根据粒子位置编码和公式(1)，计算出各粒子的适应度值；将粒子当前位置适应度值与之前迭代时该粒子的位置适应度值作比较，若前者比后者小，则更新该粒子最优位置Pi＝[pi1,pi2,…,pi(J(K+2)+1)]，否则，保持不变；同时，将该粒子对应的适应度值与之前迭代时的适应度值作比较，若前者比后者小，则更新整个粒子群的 最优位置Pg＝[pg1,pg2,…,pg(J(K+2)+1)]，否则，保持不变；Step 4: Calculate the fitness value of each particle according to the particle position coding and formula (1); compare the current position fitness value of the particle with the position fitness value of the particle in the previous iteration, if the former is smaller than the latter, Then updating the optimal position of the particle P i =[p i1 ,p i2 ,...,p i(J(K+2)+1) ], otherwise, remains unchanged; at the same time, the corresponding fitness value of the particle is The fitness values of the previous iterations are compared. If the former is smaller than the latter, the optimal position of the entire particle swarm is updated. P g =[p g1 ,p g2 ,...,p g(J(K+2)+1) ], otherwise, remain unchanged;
步骤五、若迭代次数已达到最大迭代次数，则停止迭代，对粒子群最优位置进行解码得到相应BP神经网络参数，否则进行下一步；Step 5: If the number of iterations has reached the maximum number of iterations, the iteration is stopped, and the optimal position of the particle group is decoded to obtain the corresponding BP neural network parameters, otherwise the next step is performed;
步骤六、定义粒子群的进化度σ为：Step 6. Define the evolution σ of the particle swarm as:
Figure PCTCN2015070735-appb-000010
Figure PCTCN2015070735-appb-000010
上式中，fgbest(k)和favg(k)分别为第k次迭代时粒子群的全局最优适应度值和平均适应度值；In the above formula, f gbest (k) and f avg (k) are the global optimal fitness value and the average fitness value of the particle group at the kth iteration, respectively;
定义动态变化的惯性因子计算公式为：The formula for calculating the inertia factor for dynamically changing is:
Figure PCTCN2015070735-appb-000011
Figure PCTCN2015070735-appb-000011
上式中，w0为设定的初始值。In the above formula, w 0 is the set initial value.
定义自适应变异概率计算公式为：The formula for calculating the adaptive mutation probability is:
Figure PCTCN2015070735-appb-000012
Figure PCTCN2015070735-appb-000012
Figure PCTCN2015070735-appb-000013
Figure PCTCN2015070735-appb-000013
其中，pm(k)为粒子群在第k次迭代的变异概率；pmmax和pmmin分别为变异率的最大值和最小值，fi(k)为粒子i第k次迭代时的适应度值，ε为设定的常数；Where p m (k) is the mutation probability of the particle swarm at the kth iteration; p mmax and p mmin are the maximum and minimum of the mutation rate, respectively, and f i (k) is the adaptation of the kth iteration of particle i Degree value, ε is the set constant;
定义粒子变异后新的速度和位置计算公式为：The new velocity and position calculation formula for defining particle variability is:
vid(k+1)＝w(k)vid(k)+c1r1(pid(k)-xid(k))+c2r2(pgd(k)-xid(k))   (7)v id (k+1)=w(k)v id (k)+c 1 r 1 (p id (k)-x id (k))+c 2 r 2 (p gd (k)-x id ( k)) (7)
Figure PCTCN2015070735-appb-000014
Figure PCTCN2015070735-appb-000014
其中，vid(k)为粒子i在第k次迭代中第d维的速度；Vi(k)为粒子i在第k次迭代中的速度向量；pid(k)为粒子i第k次迭代中第d维的最好位置；pgd(k)为整个群体第 k次迭代中第d维的最好位置；Xi(k)为粒子i第k次迭代中位置向量；c1和c2为非负加速常数；r1和r2为[0,1]范围内变换的随机数；R1和R2为不相等正整数，在[1,N]范围内；Where v id (k) is the velocity of the d- th dimension of the particle i in the kth iteration; V i (k) is the velocity vector of the particle i in the kth iteration; p id (k) is the particle i k The best position of the dth dimension in the iteration; p gd (k) is the best position of the dth dimension in the kth iteration of the whole group; X i (k) is the position vector in the kth iteration of the particle i; c 1 And c 2 are non-negative acceleration constants; r 1 and r 2 are random numbers transformed in the range [0, 1]; R 1 and R 2 are unequal positive integers, in the range [1, N];
步骤七、先由以上公式(4)和(5)计算出本次迭代时的惯性因子和粒子变异率，再利用公式(7)和(8)更新所有粒子的速度和位置，生成新一代粒子群，返回步骤四。Step 7. First calculate the inertia factor and particle variability of this iteration from the above formulas (4) and (5), and then update the velocity and position of all particles by using equations (7) and (8) to generate a new generation of particles. Group, return to step four.
结果，降噪前、后残余噪声测量传声器所拾取信号对比情况如图4、图5和图6所示，效果十分显著。由此可见，本实施例不仅构建了合理的单通道变压器噪声抑制系统，而且在BP神经网络滤波基础上融合了粒子群优化算法，从而利用改进粒子群优化算法改善BP神经网络中权、阈值的修正方式，克服了梯度下降优化等算法所具有的固有缺陷，能够有效地抑制变压器所有频率成分的噪声。此外，由多组单通道降噪系统组成的变压器多通道降噪系统能够有效地将变压器噪声控制在允许的范围内。As a result, the comparison of the signals picked up by the residual noise measuring microphone before and after noise reduction is shown in Fig. 4, Fig. 5 and Fig. 6, and the effect is very remarkable. It can be seen that this embodiment not only constructs a reasonable single-channel transformer noise suppression system, but also combines the particle swarm optimization algorithm based on BP neural network filtering, so as to improve the weight and threshold of BP neural network by using improved particle swarm optimization algorithm. The correction method overcomes the inherent defects of the algorithm such as gradient descent optimization, and can effectively suppress the noise of all frequency components of the transformer. In addition, a multi-channel noise reduction system consisting of multiple sets of single-channel noise reduction systems can effectively control the transformer noise within the allowable range.
以上所述，仅为本发明中的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉该技术的人在本发明所揭露的技术范围内，可理解想到的变换或替换，都应涵盖在本发明的包含范围之内，因此，本发明的保护范围应该以权利要求书的保护范围为准。 The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can understand the alteration or replacement within the scope of the technical scope of the present invention. The scope of the invention should be construed as being included in the scope of the invention.

Claims (7)

  1. 一种变压器噪声抑制方法，其特征在于：在主要由含智能芯片的控制器以及初始噪声测量传声器、残余噪声测量传声器、扬声器组成的系统中，所述控制器按以下步骤实现噪声抑制：A transformer noise suppression method is characterized in that, in a system mainly composed of a controller including a smart chip and an initial noise measuring microphone, a residual noise measuring microphone, and a speaker, the controller performs noise suppression according to the following steps:
    第一步、接收位于噪声源附近初始噪声测量传声器传输并经转换的初始噪声数字信号，作为BP(BackPropagation)神经网络的输入信号；a first step of receiving an initial noise digital signal transmitted by the initial noise measurement microphone located near the noise source and converted, as an input signal of a BP (BackPropagation) neural network;
    第二步、借助BP神经网络对输入信号进行处理后，生成相位与输入信号偏离的次级数字信号；In the second step, after the input signal is processed by the BP neural network, a secondary digital signal whose phase deviates from the input signal is generated;
    第三步、将次级数字信号转换成模拟信号经放大后输出到扬声器，产生对初始噪声具有抑制作用的次级噪声；In the third step, the secondary digital signal is converted into an analog signal, which is amplified and output to the speaker, and generates secondary noise that has an inhibitory effect on the initial noise;
    第四步、接收残余噪声测量传声器拾取的初始噪声和次级噪声叠加并经转换后的残余噪声数字信号，判断残余噪声数字信号的幅值是否已连续设定次数不变，如是则保持次级数字信号输出，如否则进行下步；The fourth step, receiving the residual noise and the secondary noise superimposed by the residual noise measurement microphone and the converted residual noise digital signal, determining whether the amplitude of the residual noise digital signal has been continuously set the number of times, and if so, maintaining the secondary Digital signal output, if otherwise proceeding to the next step;
    第五步、以使残余噪声数字信号幅值最小为优化原则，优化、调整BP神经网络参数，将上一步的残余噪声数字信号作为输入信号重复第二步。In the fifth step, the BP neural network parameters are optimized and adjusted to minimize the amplitude of the residual noise digital signal, and the residual noise digital signal of the previous step is used as the input signal to repeat the second step.
  2. 根据权利要求1所述的变压器噪声抑制方法，其特征在于：所述第五步中优化、调整BP神经网络参数采用改进粒子群算法，按以下过程进行：The transformer noise suppression method according to claim 1, wherein in the fifth step, the BP neural network parameters are optimized and adjusted by using an improved particle swarm algorithm, and the following process is performed:
    步骤一、根据已有的BP神经网络结构确定粒子的维数，并随机产生N个初始粒子；步骤二、将第n时刻神经网络中输入层神经元i与隐含层神经元h之间的权系数ωhi(n)、隐含层神经元h与输出层之间的权系数Wh(n)、隐含层神经元h的阈值GEh(n)及输出层神经元的阈值ge(n)按预定的次序进行实数编码，形成相应的实数粒子，每个粒子对应一组网络参数，其编码形式为；Step 1: Determine the dimension of the particle according to the existing BP neural network structure, and randomly generate N initial particles; Step 2, between the input layer neuron i and the hidden layer neuron h in the neural network at the nth time The weight coefficient ω hi (n), the weight coefficient W h (n) between the hidden layer neuron h and the output layer, the threshold GE h (n) of the hidden layer neuron h, and the threshold ge of the output layer neuron ( n) performing real number coding in a predetermined order to form corresponding real number particles, each particle corresponding to a set of network parameters, the coding form of which is;
    ω11 Ω 11 ω12 ω 12 …... ω1K ω 1K W1 W 1 GE1 GE 1 …... ωJ1 ω J1 ωJ2 ω J2 …... ωJK ω JK WJ W J GEJ GE J geGe
    步骤三、选择系统的残余噪声信号e(n)作为网络参数的评判标准，以如下适应度函数F(n)作为粒子位置编码和公式：Step 3: Select the residual noise signal e(n) of the system as the criterion for evaluating the network parameters, and use the following fitness function F(n) as the particle position coding and formula:
    Figure PCTCN2015070735-appb-100001
    Figure PCTCN2015070735-appb-100001
    上式中，x(n)为第n时刻输入的初始噪声数字信号；H1(z)和H2(z)分别为初级通道和次级通道的传递函数；In the above formula, x(n) is the initial noise digital signal input at the time n; H 1 (z) and H 2 (z) are the transfer functions of the primary channel and the secondary channel, respectively;
    Figure PCTCN2015070735-appb-100002
    Figure PCTCN2015070735-appb-100002
    上式中，xi(n)＝x(n-i+1)表示输入层神经元i的输入；以f(x)＝2/(1+exp(-2x))-1表示网络隐含层的激励函数；In the above formula, x i (n)=x(n-i+1) represents the input of the input layer neuron i; the network implied by f(x)=2/(1+exp(-2x))-1 The excitation function of the layer;
    步骤四、根据粒子位置编码和公式(1)，计算出各粒子的适应度值；将粒子当前位置适应度值与之前迭代时该粒子的位置适应度值作比较，若前者比后者小，则更新该粒子最优位置Pi＝[pi1,pi2,…,pi(J(K+2)+1)]，否则，保持不变；同时，将该粒子对应的适应度值与之前迭代时的适应度值作比较，若前者比后者小，则更新整个粒子群的最优位置Pg＝[pg1,pg2,…,pg(J(K+2)+1)]，否则，保持不变；Step 4: Calculate the fitness value of each particle according to the particle position coding and formula (1); compare the current position fitness value of the particle with the position fitness value of the particle in the previous iteration, if the former is smaller than the latter, Then updating the optimal position of the particle P i =[p i1 ,p i2 ,...,p i(J(K+2)+1) ], otherwise, remains unchanged; at the same time, the corresponding fitness value of the particle is The fitness values of the previous iterations are compared. If the former is smaller than the latter, the optimal position of the entire particle swarm is updated. P g =[p g1 ,p g2 ,...,p g(J(K+2)+1) ], otherwise, remain unchanged;
    步骤五、若迭代次数已达到最大迭代次数，则停止迭代，对粒子群最优位置进行解码得到相应BP神经网络参数，否则进行下一步；Step 5: If the number of iterations has reached the maximum number of iterations, the iteration is stopped, and the optimal position of the particle group is decoded to obtain the corresponding BP neural network parameters, otherwise the next step is performed;
    步骤六、定义粒子群的进化度σ为：Step 6. Define the evolution σ of the particle swarm as:
    Figure PCTCN2015070735-appb-100003
    Figure PCTCN2015070735-appb-100003
    上式中，fgbest(k)和favg(k)分别为第k次迭代时粒子群的全局最优适应度值和平均适应度值；In the above formula, f gbest (k) and f avg (k) are the global optimal fitness value and the average fitness value of the particle group at the kth iteration, respectively;
    定义动态变化的惯性因子计算公式为：The formula for calculating the inertia factor for dynamically changing is:
    Figure PCTCN2015070735-appb-100004
    Figure PCTCN2015070735-appb-100004
    上式中，w0为设定的初始值。In the above formula, w 0 is the set initial value.
    定义自适应变异概率计算公式为：The formula for calculating the adaptive mutation probability is:
    Figure PCTCN2015070735-appb-100005
    Figure PCTCN2015070735-appb-100005
    Figure PCTCN2015070735-appb-100006
    Figure PCTCN2015070735-appb-100006
    其中，pm(k)为粒子群在第k次迭代的变异概率；pmmax和pmmin分别为变异率的最大值和最小值，fi(k)为粒子i第k次迭代时的适应度值，ε为设定的常数；Where p m (k) is the mutation probability of the particle swarm at the kth iteration; p mmax and p mmin are the maximum and minimum of the mutation rate, respectively, and f i (k) is the adaptation of the kth iteration of particle i Degree value, ε is the set constant;
    定义粒子变异后新的速度和位置计算公式为：The new velocity and position calculation formula for defining particle variability is:
    vid(k+1)＝w(k)vid(k)+c1r1(pid(k)-xid(k))+c2r2(pgd(k)-xid(k))       (7)v id (k+1)=w(k)v id (k)+c 1 r 1 (p id (k)-x id (k))+c 2 r 2 (p gd (k)-x id ( k)) (7)
    Figure PCTCN2015070735-appb-100007
    Figure PCTCN2015070735-appb-100007
    其中，vid(k)为粒子i在第k次迭代中第d维的速度；Vi(k)为粒子i在第k次迭代中的速度向量；pid(k)为粒子i第k次迭代中第d维的最好位置；pgd(k)为整个群体第k次迭代中第d维的最好位置；Xi(k)为粒子i第k次迭代中位置向量；c1和c2为非负加速常数；r1和r2为[0,1]范围内变换的随机数；R1和R2为不相等正整数，在[1,N]范围内；Where v id (k) is the velocity of the d- th dimension of the particle i in the kth iteration; V i (k) is the velocity vector of the particle i in the kth iteration; p id (k) is the particle i k The best position of the dth dimension in the iteration; p gd (k) is the best position of the dth dimension in the kth iteration of the whole group; X i (k) is the position vector in the kth iteration of the particle i; c 1 And c 2 are non-negative acceleration constants; r 1 and r 2 are random numbers transformed in the range [0, 1]; R 1 and R 2 are unequal positive integers, in the range [1, N];
    步骤七、先由以上公式(4)和(5)计算出本次迭代时的惯性因子和粒子变异率，再利用公式(7)和(8)更新所有粒子的速度和位置，生成新一代粒子群，返回步骤四。Step 7. First calculate the inertia factor and particle variability of this iteration from the above formulas (4) and (5), and then update the velocity and position of all particles by using equations (7) and (8) to generate a new generation of particles. Group, return to step four.
  3. 根据权利要求2所述的变压器噪声抑制方法，其特征在于：所述智能芯片为DSP芯片，所述初始噪声测量传声器和残余噪声测量传声器分别经过自动增益调节器的声音处理芯片和A/D转换芯片后与DSP芯片对应的输入端口连接，所述DSP芯片的输出端通过D/A转换芯片和功率放大器后接扬声器。The transformer noise suppression method according to claim 2, wherein the smart chip is a DSP chip, and the initial noise measuring microphone and the residual noise measuring microphone are respectively subjected to a sound processing chip and an A/D conversion of an automatic gain adjuster. The chip is connected to an input port corresponding to the DSP chip, and the output end of the DSP chip is connected to the speaker through a D/A conversion chip and a power amplifier.
  4. 根据权利要求3所述的变压器噪声抑制方法，其特征在于：所述BP神经网络结构设为2-10-1滤波算法，其中隐含层和输出层的激励函数分别为Sigmoid函数和Purelin函数。The transformer noise suppression method according to claim 3, wherein the BP neural network structure is set to a 2-10-1 filtering algorithm, wherein the excitation functions of the hidden layer and the output layer are respectively a Sigmoid function and a Purelin function.
  5. 根据权利要求4所述的变压器噪声抑制方法，其特征在于：所述粒子群的规模为30，最大遗传代数为10，所述粒子的维数为78。The transformer noise suppression method according to claim 4, wherein the particle group has a size of 30, a maximum genetic algebra of 10, and a dimension of the particle of 78.
  6. 根据权利要求5所述的变压器噪声抑制方法，其特征在于：The transformer noise suppression method according to claim 5, wherein:
    所述初始噪声测量传声器距变压器10cm，其高度位于油箱中部；扬声器距初始噪声测量传声器20cm，其高度位于油箱中部；残余噪声测量传声器距扬声器20cm，其高度位于油箱中部。上述三种器件按顺序排成一列，构成单通道降噪系统。 The initial noise measuring microphone is 10 cm away from the transformer, and its height is located in the middle of the fuel tank; the speaker is 20 cm away from the initial noise measuring microphone, and its height is located in the middle of the fuel tank; the residual noise measuring microphone is 20 cm from the speaker, and its height is located in the middle of the fuel tank. The above three devices are arranged in a row to form a single channel noise reduction system.
  7. 根据权利要求5所述的变压器噪声抑制方法，其特征在于：The transformer noise suppression method according to claim 5, wherein:
    所述变压器多通道降噪系统由8个单通道降噪系统组成。在变压器正和反两面均放置3组单通道降噪系统，其位置分别与变压器的A、B、C三相对应；在变压器测面的中间位置放置1组单通道降噪系统。 The transformer multi-channel noise reduction system consists of eight single-channel noise reduction systems. Three sets of single-channel noise reduction systems are placed on both the positive and negative sides of the transformer, and their positions correspond to the three-phase A, B and C of the transformer respectively; one set of single-channel noise reduction system is placed in the middle of the transformer measuring surface.
PCT/CN2015/070735 2014-05-06 2015-01-15 A transformer noise suppression method WO2015169109A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US15/323,407 US10083709B2 (en) 2014-05-06 2015-01-15 Transformer noise suppression method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410189040.5A CN103971908B (en) 2014-05-06 2014-05-06 A kind of transformer noise suppressing method
CN201410189040.5 2014-05-06

Publications (1)

Publication Number Publication Date
WO2015169109A1 true WO2015169109A1 (en) 2015-11-12

Family

ID=51241291

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/070735 WO2015169109A1 (en) 2014-05-06 2015-01-15 A transformer noise suppression method

Country Status (3)

Country Link
US (1) US10083709B2 (en)
CN (1) CN103971908B (en)
WO (1) WO2015169109A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107734433A (en) * 2017-09-08 2018-02-23 中国飞行试验研究院 Based on the star solid microphone array optimization method for improving particle cluster algorithm
CN108733924A (en) * 2018-05-21 2018-11-02 东南大学 A kind of intellectualized design method of digital coding metamaterial unit

Families Citing this family (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103971908B (en) * 2014-05-06 2016-03-09 国家电网公司 A kind of transformer noise suppressing method
CN104821166A (en) * 2015-01-10 2015-08-05 哈尔滨工业大学（威海） Active noise control method based on particle swarm optimization algorithm
CN106340290A (en) * 2016-11-09 2017-01-18 国家电网公司 Active noise reduction method and device
CN107070802B (en) * 2016-12-21 2021-01-12 吉林大学 Wireless sensor network congestion control method based on PID controller
KR20180111271A (en) * 2017-03-31 2018-10-11 삼성전자주식회사 Method and device for removing noise using neural network model
CN109036810B (en) * 2017-04-18 2020-12-29 江西明正变电设备有限公司 Transformer active noise reduction device and control method
CN107219763A (en) * 2017-06-07 2017-09-29 安徽师范大学 A kind of secondary channel line modeling method declined based on gradient
CN107240404B (en) * 2017-06-08 2020-05-01 中国电建集团福建省电力勘测设计院有限公司 Noise reduction method for prefabricated cabin type transformer substation
CN107393545B (en) * 2017-07-17 2020-12-08 会听声学科技（北京）有限公司 Feedback type active noise reduction system and method with flexible gain
WO2019026973A1 (en) * 2017-08-04 2019-02-07 日本電信電話株式会社 Signal processing device using neural network, signal processing method using neural network, and signal processing program
CN107768119B (en) * 2017-11-01 2020-12-11 国网山西省电力公司电力科学研究院 Active noise reduction system of extra-high voltage power transformer
CN107871157B (en) * 2017-11-08 2020-06-09 广东工业大学 Data prediction method, system and related device based on BP and PSO
CN108593093B (en) * 2018-03-14 2021-07-02 中国电力科学研究院有限公司 Ultra-high voltage transformer acoustic power field measurement method
CN108646598B (en) * 2018-05-04 2022-02-08 南京大学 Low-order analog feedback controller suitable for miniaturization and design method thereof
CN110709921A (en) * 2018-05-28 2020-01-17 深圳市大疆创新科技有限公司 Noise reduction method and device and unmanned aerial vehicle
CN109856969B (en) * 2018-11-06 2023-10-03 皖西学院 Fault prediction method and prediction system based on BP neural network model
CN110008633B (en) * 2019-04-19 2023-04-14 中国石油大学(华东) Highway noise suppression method and system based on artificial intelligence deep neural network
KR102263250B1 (en) * 2019-08-22 2021-06-14 엘지전자 주식회사 Engine sound cancellation device and engine sound cancellation method
KR102225975B1 (en) * 2019-08-26 2021-03-10 엘지전자 주식회사 Engine sound synthesis device and engine sound synthesis method
CN111091805B (en) * 2019-11-15 2023-05-26 佳禾智能科技股份有限公司 Feedback type noise reduction method based on neural network
CN111063528B (en) * 2019-11-25 2021-08-31 深圳供电局有限公司 Active noise reduction equipment of transformer and transformer noise on-line monitoring system
CN111640415B (en) * 2020-06-03 2023-04-07 吉林大学 Active noise reduction algorithm parameter optimization method based on particle swarm optimization
CN111899710A (en) * 2020-07-10 2020-11-06 广东电网有限责任公司 Method and device for actively reducing noise of power distribution room based on analytic hierarchy process
CN112151249B (en) * 2020-08-26 2024-04-02 国网安徽省电力有限公司检修分公司 Active noise reduction method and system for transformer and storage medium
CN112163274B (en) * 2020-09-25 2021-08-06 中车长春轨道客车股份有限公司 Low-noise optimization design method and system in vehicle
CN112365074B (en) * 2020-11-18 2023-08-18 贵州电网有限责任公司 Artificial intelligence decision-making method based on power grid regulation and control data
CN112665581A (en) * 2020-12-04 2021-04-16 山东省计算中心（国家超级计算济南中心） Combined navigation method based on BP neural network assisted Kalman filtering
CN112727704B (en) * 2020-12-15 2021-11-30 北京天泽智云科技有限公司 Method and system for monitoring corrosion of leading edge of blade
CN112785522B (en) * 2021-01-21 2024-02-20 国网山西省电力公司电力科学研究院 Self-adaptive PCNN image denoising method based on GWO and BEMD optimization
CN112883628A (en) * 2021-03-23 2021-06-01 上海交通大学 Method and system for positioning abnormal sound source of transformer substation equipment
CN113076516A (en) * 2021-04-16 2021-07-06 上海欣影电力科技股份有限公司 PSO-EMMD-ICA-based transformer vibration signal online separation method
CN113630098B (en) * 2021-08-11 2023-08-15 深圳远虑科技有限公司 Gain control method and device of audio amplifier and electronic equipment
CN113963713A (en) * 2021-10-11 2022-01-21 稿定（厦门）科技有限公司 Audio noise reduction method and device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102008061552A1 (en) * 2008-12-11 2010-07-01 Areva Energietechnik Gmbh Method for reducing noise of electrical transformer, involves determining current operating point of transformer and providing measurement protocol for characterizing operating point dependent behavior of transformer
CN102208263A (en) * 2011-02-24 2011-10-05 南京大学 Multi-channel active noise control system for power transformer
CN103475336A (en) * 2013-09-06 2013-12-25 深圳供电局有限公司 Power transformer noise control method based on inverse control technology
CN103971908A (en) * 2014-05-06 2014-08-06 国家电网公司 Transformer noise suppression method

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3693749A (en) 1971-04-26 1972-09-26 Gen Electric Reduction of gas turbine engine noise annoyance by modulation
US5293578A (en) * 1989-07-19 1994-03-08 Fujitso Ten Limited Noise reducing device
US5315661A (en) * 1992-08-12 1994-05-24 Noise Cancellation Technologies, Inc. Active high transmission loss panel
US5381485A (en) * 1992-08-29 1995-01-10 Adaptive Control Limited Active sound control systems and sound reproduction systems
US5386689A (en) * 1992-10-13 1995-02-07 Noises Off, Inc. Active gas turbine (jet) engine noise suppression
US6763339B2 (en) * 2000-06-26 2004-07-13 The Regents Of The University Of California Biologically-based signal processing system applied to noise removal for signal extraction
AU2005100274A4 (en) * 2004-03-31 2005-06-23 Kapur, Ruchika Ms Method and apparatus for analyising sound
JP2009096259A (en) * 2007-10-15 2009-05-07 Fujitsu Ten Ltd Acoustic system
CN102013090A (en) * 2010-11-23 2011-04-13 电子科技大学 Passive millimetre wave image strip noise suppression method
CN103971988A (en) 2014-05-26 2014-08-06 宁波黎明继电器有限公司 Pressure relay

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102008061552A1 (en) * 2008-12-11 2010-07-01 Areva Energietechnik Gmbh Method for reducing noise of electrical transformer, involves determining current operating point of transformer and providing measurement protocol for characterizing operating point dependent behavior of transformer
CN102208263A (en) * 2011-02-24 2011-10-05 南京大学 Multi-channel active noise control system for power transformer
CN103475336A (en) * 2013-09-06 2013-12-25 深圳供电局有限公司 Power transformer noise control method based on inverse control technology
CN103971908A (en) * 2014-05-06 2014-08-06 国家电网公司 Transformer noise suppression method

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107734433A (en) * 2017-09-08 2018-02-23 中国飞行试验研究院 Based on the star solid microphone array optimization method for improving particle cluster algorithm
CN107734433B (en) * 2017-09-08 2021-11-02 中国飞行试验研究院 Star-shaped stereo microphone array optimization method based on improved particle swarm optimization
CN108733924A (en) * 2018-05-21 2018-11-02 东南大学 A kind of intellectualized design method of digital coding metamaterial unit
CN108733924B (en) * 2018-05-21 2022-05-10 东南大学 Intelligent design method of digital coding metamaterial unit

Also Published As

Publication number Publication date
CN103971908B (en) 2016-03-09
US20170249955A1 (en) 2017-08-31
CN103971908A (en) 2014-08-06
US10083709B2 (en) 2018-09-25

Similar Documents

Publication Publication Date Title
WO2015169109A1 (en) A transformer noise suppression method
Pawelczyk Adaptive noise control algorithms for active headrest system
CN103474060B (en) A kind of power equipment Noise Active suppressing method based on internal model control
CN103841497B (en) The method of the more new algorithm and decorrelation unit of the adaptive feedback estimating system of control
CN104935293B (en) High-power transformer adaptive active method for noise reduction control and control system
CN103035234B (en) Active noise reducing device and method and the noise control system for smoke exhaust ventilator
CN101901602B (en) Method for reducing noise by using hearing threshold of impaired hearing
CN103534942B (en) Process audio signal
US9443533B2 (en) Measuring and improving speech intelligibility in an enclosure
CN105706466A (en) Hearing aid with probabilistic hearing loss compensation
CN109346052B (en) Device and method for optimizing sound quality in vehicle by utilizing active noise reduction
CN103841496A (en) Self-adaption indoor transformer noise elimination device and implement method thereof
CN106340290A (en) Active noise reduction method and device
George et al. A reduced complexity adaptive Legendre neural network for nonlinear active noise control
CN106162482A (en) A kind of loudspeaker array sound reinforcement system uttered long and high-pitched sounds for suppression and method
EP2641346B2 (en) Systems and methods for reducing unwanted sounds in signals received from an arrangement of microphones
CN108710106A (en) A kind of active head rest device based on infrared positioning
Patel et al. Partial update even mirror fourier non-linear filters for active noise control
CN109147753B (en) Convex combination noise reduction method based on minimum difference between square error and square logarithm error
Sharma et al. Server noise: health hazard and its reduction using active noise control
CN111462724A (en) Noise self-adaptive noise reduction method for high-voltage converter station
Nakrani et al. Feed-Forward and feedback active noise control system using FxLMS algorithm for narrowband and broadband noise
CN113096629A (en) Relative path virtual sensing method for single-channel feedback active noise control system
CN105744456A (en) Digital hearing-aid self-adaptive sound feedback elimination method
de Diego et al. On the performance of a local active noise control system

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15788861

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 15323407

Country of ref document: US

122 Ep: pct application non-entry in european phase

Ref document number: 15788861

Country of ref document: EP

Kind code of ref document: A1